W^^^^^^SSgg^^^mmml Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium IVIember Libraries http://www.archive.org/details/revealedperformaOOchir (}A. Massachusetts Institute of Technology Department of Economics Working Paper Series A Revealed Preference Approach to Understanding Corporate Governance Problems: Evidence from Canada Robert S. Chirinko Huntley Schaller Working Paper 02-46 November 2002 Room E52-251 50 Memorial Drive Cambridge, MA 02142 This paper can be downloaded without charge from the Social Science Research Network Paper Collection at http://ssm.com/abstract_id=359 1 80 MASSACHUStnS IMSTITUTE OF TECHNOLOGY ) A Revealed Preference Approach To Understanding Corporate Governance Problems: Evidence From Canada Abstract Empire -building by managers implies that they use a lower effective discount rate in making investment decisions. We use actual investment decisions to measure the gap between the manager's effective discount rate and the market rate. Our empirical work is based on panel data for 193 Canadian fimis. Distinctive institutional features, such as interrelated groups of Canadian firms and concentrated share ownership, allow us to quantify the sensitivity of effective discount rates and governance problems to these institutional control mechanisms. For the firms most be affected by the agency problems highlighted by Jensen (1986), estimated discount rates are 350-400 basis points less than the market rate, supporting the Free likely to Cash Flow view govemance problems distort finn Flow problems have a stock of 22% higher than would prevail under value that unresolved corporate behavior. Finns in our sample that face Free Cash fixed capital approximately 7% to maximizing behavior. E2 Keywords Corporate Govemance, Business Investment, Discount Rates TF.T.Nos: G3, : Corresponding Author : Professor Robert S. Chirinko Professor Huntley Schaller Department of Economics Department of Economics Emory Carleton University University Atlanta, Georgia USA 1 125 Colonel By 30322-2240 PH: (404) 727-6645 FX: (404) 727-4639 Ottawa, Ontario EM: rchirin@emory.edu EM: PH: FX: Drive K1S5B6 CANADA 520-2600 (x-375 1 (6 1 3) 520-3906 (6 1 3) schaller@ccs.carleton.ca . Table Of Contents 1 . Introduction 2. The Net Present Value Rule And The Euler Investment Equation 3. The Econometric Equation And Estimation Strategy 1 Variables In The Econometric Equation 2. The Discount Factor And The Estimation Strategy 4. The Canadian Dataset 1 Financial Statement Data 2. The Market Discount Rate 3. Summary Statistics 5. Assessing Corporate Governance Problems 6. Curing Free Cash Flow Problems? 7. Capital Structure 8. Summary And Conclusions And Discount Rates References Appendix A: Departures From Value Maximization Appendix B: The Real Risk-Adjusted Market Discount Rate: Variable Construction And Data Sources Tables 1 A Revealed Preference Approach To Understanding Corporate Governance Problems: Evidence From Canada 1. Introduction This paper proposes a "revealed preference" approach for examining corporate governance problems in terms of discount rates. As established by many theories, governance problems have a direct and immediate impact on the effective discount rate guiding investment decisions made by executives. Our approach uses actual investment spending on business plant and equipment, and uncovers the discount rates guiding investment behavior. By comparing executives and shareholders, we which governance problems distort firm behavior. Our generate new the discount rates used evidence quantifying the extent to between discount analysis distinguishes by rates used by shareholders and those used by corporate executives. Shareholders' discount rates have been studied by many researchers usually in terms of the relations and earnings." As noted by Stein (1996) and shareholders' discount rates have corporate executives. The latter among stock prices, dividends, Dow and Gorton (1997), biases in no necessary implications for those used by discount rates are critical for determining investment spending, and hence are key to understanding governance issues and possible distortions of firm behavior. This study generates evidence bearing on governance problems by examining actual patterns of investment spending to reveal the effective discount rate used executives. Govemance problems arise by from divergent incentives and asymmetric information between owners and executives. These conflicts, coupled with the impossibility of writing contracts covering all friture contingencies, lead to unresolved For example, see Daniels and Morck (1995) for Canada, Hall and Hall (1993) for the United and references cited therein. States, 2 governance problems affecting finn behavior (Hart, 1995). The Free Cash Flow model of Jensen (1986) emphasizes that executives those of owners and pursue projects that, when have evaluated interests that differ at the from market discount rate used by shareholders, have negative net present value (T^PV). Such a departure from value maximization can be interpreted in terms of an effective discount rate that low While acknowledging the relative to the shareholders' discount rate. difficulties caused by agency conflicts, an altemative set is potential of theories holds that pressures operating through the capital, product, and factor markets are sufficiently strong to largely overcome these problems.^ From the perspective of this Perfect Markets model, financial markets allocate resources efficiently and investment decisions and discount rates remain unbiased by potential governance problems. ]n this paper, light we compare executives' and shareholders' discount rates in order to shed on these theories of corporate governance. The econometric equation in the next capital two that is the basis for our empirical work is sections. Section 2 begins with the net present value rule developed used in budgeting and, after suitable fransformation, develops an estimating equation that links investment spending, discount rates, estimating equation is and govemance problems. This equivalent to the Euler investment equation. Section 3 discusses specification issues, and shows how information from the fransformed net present value rule and variation in frrm-level panel data can "reveal" the effective discount rate guiding investment decisions. depreciation, we With adjustments for risk estimate the gap between the discount rates used and by executives and shareholders by sorting firms into mutually exclusive and exhaustive confrasting ^ See Alchian (1950), Demsetz and Lehn (1985), and Miller (1997). Relying on legal protections outcome of competition among states) to defend the interests of owners, Carney (1997), Easterbrook and Fischel (1991), and Romano (1993) reach a similar conclusion about the corporate govemance system in the United States. (possibly determined as the on firm classes based characteristics identified in the literature as indicating governance problems (e.g., firms with high free cash flow and poor investment opportunities). Section 4 discusses the sources and construction of the variables in our panel dataset containing 193 Canadian firms. Canada is a useful "laboratory" in which to study corporate governance in light of the recent challenge to the Berle-Means perspective advanced by La and Shleifer (1999). They Porta, Lopez-de-Silanes, argue that the Berle-Means firm with widely-dispersed ownership rather than the rule for industrialized exceptions are the United is the exception and industrializing economies. The notable Kingdom and the United States where 90% and 80%, These prototypical Berle-Means respectively, of the largest firms are widely held. economies contrast with a comparable sample average of 24%." However, the Canadian economy occupies an exceedingly useful middle ground with 50%) of large firms being widely held. Consequently, there variation with which to study corporate is a notable govemance amount of institutional issues. The existence of concentrated share ownership in Canada and interrelated groups of Canadian frrms allow us to assess the sensitivity of the executives' discount rates to these distinctive institutional structures. Empirical results are contained in the next two sections. In Section the Perfect Markets model, but find strong support for the Free "Jensen firms" ~ those firms most flow discussed by Jensen (1986) likely to suffer statistical significance ability reject Cash Flow model. discount rates 350-400 basis by shareholders. With the economic and of Free Cash Flow problems established. Section 6 assesses the of various mechanisms The ownership we from the agency costs of free cash ~ have risk-adjusted points lower than the discount rates used 5, figures are from to attenuate these La problems. Porta, Lopez-de-Silanes, We document that and Shleifer (1999, Table II. B). 4 concentrated ownership is associated with higher (less distorted) discount rates for Jensen firms. Section 7 apphes our procedure for estimating effective discount rates to capital structure issues. In a recent survey that a unified theory lists the Free of the literature, of capital structure applicable to Cash Flow and Pecking Order models Does our evidence favoring the Free firms has not emerged, and all as Myers (2001) emphasizes two of the leading Cash Flow model, preclude theories. the empirical importance of the financing problems highlighted by the Pecking Order model? measure the distortions associated with these estimated fi-om a common models in fi-amework, and find that, for We terms of discount rates some firms at some times, Free Cash Flow problems are important, while at other times, perhaps for the same firm, Pecking Order problems are important. Section 8 presents a summary and shareholders' discount rates, conclusions. By comparing executives' and we document that govemance problems of the sort emphasized by Jensen's Free Cash Flow model are substantial and reduced by the close monitoring afforded by concentrated ownership. The firms identified in this study is approximately capital stock of Jensen 7% to 22% too large because of govemance problems. 2. The Net Present Value Rule And The Euler Investment Equation The general principle guiding our estimation strategy is to infer discount rates used by corporate executives fi^om the patterns of their actual investment spending. The estimating equation is based on the net present value (NPV) rule used budgeting that links investment spending, discount problems. This section shows how rates, and corporate govemance a simple transformation of the NPV mle generates the Euler equation usually derived fi-om the standard intertemporal investment spending found in the economics literature. in capital model of real 5 We begin with the fundamental capital budgeting decision rule that identifies positive NPV projects by comparing the marginal costs and benefits from acquiring an additional unit of capital, 00 S MIC, <_ R= ICFt+j R\ (1) l/(l+r), where MIC, is the (2) sum of the purchase price of a new unit price of output and the marginal adjustment cost at time t, of capital relative to the the ICFi+j's are the incremental cash flows per unit of capital associated with the project that will accrue in current and future periods, and real risk-adjusted discount rate, ( 1 ) is R is the real The r. discount factor defined in terms of the firm should continue to acquire capital until satisfied with equality. To assuming derive the Euler investment equation from the marginal investment equates benefits and costs, and hence (1) can that the be analyzed as an equality for period S ICF,+j RJ = ICF,+ j=0 Consider the both sides by t as follows, 00 00 MIC, = NPV mle, we begin by S ICF,+j R^ (3) j=l NPV rule in period t+1 R and redefine the j and, to facilitate eventual cancellation, multiply index appropriately. For expositional convenience, the derivations in this section are based on the assumptions of a constant discount rate and non-depreciating capital. TTiese assumptions can be relaxed if the estimating equation is derived from a dynamic optimization problem with variational methods. See Chirinko (1993) for a formal analysis of the Euler equation. Risk-adjustment 4.2 and Appendix B. is discussed in Section OO R * MIC,+, 00 E = R S = ICF,+i+j RJ j=0 - (MICt Equation (5) is - R * (4) j=l All future ICF's can be eliminated ICFt ICF.^j R^. = MIC,+,) by subtracting (4) from (3) and rearranging, 0. (5) the Euler investment equation describing optimal investment behavior and the intertemporal tradeoff between costs and benefits of acquiring an additional unit of capital. To obtain an intuitive understanding of the Euler equation, assume that the firm is currently on the path of optimal capital accumulation. the firm will be indifferent to an increase in capital of 1 by 1 unit in period unit in t+1, thus leaving the capital stock unaffected The benefit of this stock is costly, hypothetical change represented and the Euler investment equation investment cost incurred in t remain on its sets by t now need to this path, and a decrease from period t+1 onward. ICFt. Changing the capital ICFt equal to the marginal minus the marginal investment cost saved saving arises because the firm does not capital to is Along in t+1 (this acquire an additional unit of optimal accumulation path). The t+1 saving is placed on a present value basis by the firm's one-period discount factor, R. 3. The Econometric Equation And Estimation Strategy 3. 1. Variables In The Econometric Equation The econometric equation used rule. In order to estimate (5), we need to subscript indexes firms. MlCj^t 4. 1 .) in estimation is and marginal adjustment is the costs. based on the transformed specify MICi,t, ICFjt, and sum of purchase costs (p i.t, R,,t, NPV where the i discussed in Section We assume (as is standard in the literature) that marginal adjustment costs depend on the investment/capital ratio (INVTi,t), and are . defined by a second-order Taylor expansion. = MICi.t p,,, + + ailNVT,,, + a2lNVT,,, 2x). (oo (6) All variables are adjusted for taxes. With respect to the ICF,,t, assume for the moment that the firm's production function exhibits constant returns to scale, product markets are perfectly competitive, and adjustment costs are absent. In this special case, the marginal product equals the average product of capital, where the latter divided by the capital stock. total variable costs, all is of capital defined as total revenues less We adopt less restrictive assumptions by allowing the production function to be homogeneous of degree (where ^ is t, not necessarily equal to unity, the value defining constant returns to scale) and the product markets to be imperfectly competitive. Rewriting the revenue function (which reflects both production and adjustment costs), using Euler's Theorem on Homogeneous Functions, and rearranging terms product of capital, = ICFi,t we v|/*REV,,t and returns to scale \j/ > 1 y is - COST,,t - (oo + ailNVTi.t a parameter capturing the = 0), and 4. 1 y (7) equates marginal Lastly, an error term See Section costs, respectively, divided combined effects for details (e,,.t) is (7) by the of non-constant to scale implies that returns to scale or non-competitive product markets, In the special case described above, (the a's + a2lNVTi/)*INVTi,, and imperfect competition. Increasing returns Under decreasing . marginal obtain the following specification, where REVi,t and COST,,t are revenue and variable capital stock, to isolate the y< 1 equals one, marginal adjustment costs are zero and average products of capital. added to the about tax adjustments. Euler equation.^ These considerations , 8 lead to the following estimating equation that - ICF,,, where 3.2. (MIC,,, Rj,t is R,, * MIC,,t+i) - = is the basis for the empirical results, (8) e,,, the discount factor to be discussed in the next sub-section. The Discount Factor And The Estimation Strategy The discount factor contains three components: the shareholders' discount rate c (r i,t, equal to the risk-adjusted real market interest rate), a firm-specific geometric depreciation rate (5i), used by executives (r and the gap (possibly equal to zero) between the discount i,t) rates and shareholders. To further motivate our revealed preference approach with its focus on discount rates, Appendix parsimonious model showing A contains a how utility-maximizing behavior by managers with a preference for empire-building implies that the discount rate used by executives will be less than that The used by shareholders. This gap is represented by a parameter, quantitative impact of governance problems on firm behavior by comparing the investment behavior of contrasting classes of firms is ^. assessed and using that information to infer the discount rates being used by executives. Classes are defined by characteristics identified previously in the literature as indicating governance problems. For example, the governance problems emphasized by Jensen for firms with high fi^ee these characteristics criteria cash flow and poor investment opportunifies. Since some of may change we for a given firm over time, we evaluate the sorting recognize the possibility that factors other than governance If the Euler equation is derived is occur and form our contrasting classes for each firm/year observation. In computing the estimates, term may from a formal dynamic optimization problem, this stochastic error linked directly to expectation errors and technology shocks. With depreciating capital, the saving in (1-5) units of surviving capital. marginal investment costs in period t+1 is lowered to only 9 problems may affect discount and capture the impact of these factors by rates, including an additional term in the discount rate, common (^, to all firms in the sample. With the sorted samples, distortions due to governance problems are measured by the parameter particular (j), which estimated only for the class associated with a is governance problem. These considerations lead to the following formulation of the discount factor. Ri,, = (j) where (1-6,) / ((l+r\t)(l+C + = (r\t-r\,). Tu^ is (9b) an indicator variable free cash flow — 1 if a firm falls into a particular class and poor investment opportunities), systematic risk and depreciation (which differ affecting (9a) r,,t*(t))). firms, our all among fmns) and arise correlated with the error temi. because of measurement or specification mental variables estimation technique Moments technique (GMM) because is it required. error. An (j), factor the gap contemporaneous and additional correlation Consequently, an instru- We use the Generalized Method of delivers consistent parameter estimates and allows for a general pattern of conditional heteroscedasticity See Hansen (1982). The instruments are a constant, Ri,.-i(I-8,)p',,t-i, is common rates. Lastly, regarding estimation, the Euler equation contains might the measure of corporate governance problems may be high otherwise. Conditional on between the executives' and shareholders' discount future variables that (e.g., REVi,t.i, Ri,,-i(l-5,)INVT,.-i, R,,,-i(l-5,)INVTV,,time (t), All variables are adjusted for taxes as discussed in Section 4. 1 . in the residuals. COSTj.n, ENVT.t-i, INVT',.t-i, p i,t-i, time squared (t"), and Tu (cf. (10)). The instruments values of the terms in the Euler equation (8) and the auxiliary relations ((6), (7), are primarily lagged and (9)). Such instruments are chosen because they are valid (uncorrelated with the error term) and relevant (correlated with the variables appearing in the regression). We also use a constant time, and time squared, which are attractive instruments because they are unequivocally exogenous. difficult to estimate ao; instead, we It proved undertake a grid search, choosing the value ao that minimizes the 10 4. 4. 1 The Canadian Dataset . Financial Statement Data Our empirical work is based on a balanced panel of 1 93 Canadian firms for the period 1973 to 1986. Various computations and the presence of variables dated t+1 in The the estimating equation reduce the period used in estimation to 1975 to 1985. basic sources are the are CANSIM, Laval, and Financial Post databases. comparable to the often used CRSP The latter two COMPUSTAT databases, respectively, and for U.S. finns. The estimating equation (8) requires data for investment, revenues, costs, the physical capital stock. Three series are drawn directly Annual Corporate Database. Investment (INVTi.t) property, plant and equipment. Revenue (REVjt) is is from the Financial Post gross capital expenditures on net sales. Cost (COSTi,t) revenue minus operating income, where operating income the principal activities of a business after deducting and all is the is income derived from operating expenses except depreciation, depletion and amortization, interest expense and miscellaneous expenses included in the other income/expense category. All three series are divided by the current dollar replacement cost of the fixed capital stock (K$i,t). This capital stock is constructed for a given firm in two steps. estimates the depreciation rate The first step We calculate firm-specific depreciation rates (5,). based on the firm's reported depreciation. The details are based on the procedure used by Salinger and Summers (1983). which evolves as K$i,t = (K$j,t-i+Ii,t) gross capital expenditures and machinery and equipment The tax-adjusted GMM criterion. A recursive formula is then used to calculate K$i,t, p' t is * (p' t/p' t-i) * (l-S,), where Ii,t is current dollar the implicit price index for business investment in (CANSIM series relative price ratio, Dl p i,t, 1 123). also enters the estimating equation, 11 and equals * (p' i/p°t). and provincial corporate income tax rates (( l-itCt-T,,tZ,)/(l-T|,,)) industries), the investment tax credit rate depreciation allowances domestic demand (z,). (CANSIM The output senes 1 1 1 The tax terms capture the sum of federal incorporating variation across time and (T,,t, and the present value of tax (itc,), price (p"t) is the imphcit price index for final 30). Variables defining the contrasting classes will be discussed below. 4.2. The Market Discount Rate The discount factor defined in (9)) depends on, inter alia, the (R|,t, shareholders' real risk-adjusted discount rate, n,t = r ,.i. defined as follows: ((1+1,0 /(l+Tt'O)- 1.0. (10) This rate discounts annual cash flows fi^om the middle-of-period period t+1 (See Appendix . data sources.) inflation rate is B to the middle-of- for specific details concerning variable construction The components of r represented by 7i\. j,i are defined as follows. which equals actual operating in the interests of shareholders, by t it and The one year expected inflation. If the firm is will use the nominal discount rate defined i,,t, i,,t = m-Tu)i"'^\ + where h is (l-X,)i^^^^-,,„ the firm-specific leverage ratio, provincial) of corporate income Paper is i\, rate, = 1^ (11) and + 0,, i ,,t taxation, i t,,i is i the marginal rate (federal is the nominal, one year. and Commercial the nominal, short-term, risk-adjusted cost of equity capital, (12) 12 where F i t is the nominal, premium estimated by Appendix 4.3. B one year either the risk-free Treasury Bill rate, and o, is the equity risk CAPM or the Fama-French three-factor model (see for details). Summary Statistics Summary the full sample. statistics are presented in Table Columns 2-3 and 4-5 are sorted 1 for all 193 firms. by Column 1 is for characteristics defining Jensen firms that will feature prominently in the empirical results. 13 5. Assessing Corporate Governance Problems In a classic paper, Jensen (1986) argues that executives are squander resources on negative more likely to NPV projects when a firm has substantial He tlow and poor investment opportunities. identified the oil industry free cash from 1973 the mid-1980's as being particularly susceptible to these precondifions. to We therefore begin by comparing the investment behavior and discount rates used by firms in the oil to industry to the remaining firms. A firni is classified as an "oil finn" if Our Toronto Stock Exchange (TSE) industry 300. discount rate used by oil industry executives shareholders' discount rate at the 10% level. (r estimates suggest that the and r r is statistically These resuhs are consistent with Jensen's conjecture problems and overinvestment oil CAPM in column 2. significant that, during finns suffered from unresolved governance in fixed capital. specifications reported in risk-adjustment, either the belongs about 260 basis points lower than the The gap between ). our estimation period 1975-1985, The is it row Fama-French 1 of Table 2 differ only by the method of model (FF3) three-factor in column 1 The gap between executives' and shareholders' discount very similar regardless how we Based on FF3, adjust for risk. oil or the rates is industry executives use a discount rate 260 basis points lower than that used by shareholders; based on CAPM, the comparable figure is 280 basis points. adjustment yield estimates of the gap between of each other. This F estimates of r is r Thus and r the two methods of risk that are within 20 basis points a general finding throughout our empirical results: the *s - r based on the FF3 are always close to those based on the never diverging by more than one standard primarily on the results based on the FF3 error. CAPM, The following discussion will focus risk adjustment. Table 2 also displays estimates of y, the parameter capturing deviations from Q In terms of equation (9), r equals one for oil firms; zero for all other fums. 14 constant returns to scale in production and perfect competition in the product market. This parameter of \|/ is estimated precisely, and less than unity is consistent its value is always A value less than unity. with decreasing returns to scale regardless of the competitive nature of output markets or increasing retums and a sufficient degree of imperfect competition to force the marginal return to capital below In either case, in the model can exist in it in its average return. implies that the firms are earning positive economic rents. Appendix A, such equilibrium only rents are important because if the As shown governance problems firm earns positive economic profits that allow it to "indulge" in sub-optimal investment. A second way of identifying Jensen-type governance problems is to focus on those firms with high flow (FCFi) interest is fi'ee cash flow and poor investment opportunities. Free cash defined as cash flow (revenues less cash expenses including tax and payments) less dividends, all for the previous year, divided replacement value of fixed capital. by the Investment opportunities are measured by the Brainard-Tobin Q.'° With these definitions, contrasting classes are formed in a threestep process. First, 2/3 of their for which TSE we find the firm/year observations for industry in a given year. Second, Q is in the bottom 2/3 of their TSE form the intersection of these two sets sample) to identify Jensen firms. As shown '" we which FCFi we industry in a given year. Third, (which contains approximately one-half of the in row 2 of Table 2 - is in the top find the firm/year observations for the Q is defined for the beginning of the period as follows, Q = [(E + D TDB) [K$(1-5)(1-t)], where E is - FF3 risk K$( 1-6)7] the market value of the firm's equity at the beginning of period market value of existing short-term and long-term debt, t, / D is the TDB is the tax depreciation bond representing the present value of all tax depreciation allowances claimed in the current period on K$ is the current dollar replacement cost of fixed capital, 6 is the T equals (l-itc-T*z) representing adjustments for the investment tax existing capital (Hayashi, 1982), rate of economic depreciation, and tax depreciation (t*z) per Section 4.1, and t is the corporate income tax rate reflecting federal and provincial tax rates. The TDB series (as well as the present value of credit (itc) depreciation allowances Salinger and Summers on a unit (1983). of new capital) is calculated using the method suggested by 15 adjustment, our estimate suggests that the executives of firms with high free cash flow and poor investment opportunities use a discount rate discount rate of shareholders. This gap between r As a check of the robustness of our results, E 350 basis points lower than the and we S r is statistically significant. also consider a second defined by tree cash flow in the previous two years (FCF2). row 3 of Table 2 indicate that the gap between E r S and r The measure results displayed in remains economically and statistically significant. The above sortings are based on preconditions governance problems associated with constructed based on a fi-ee for the An cash flow. emergence of additional test can be symptom of governance problems. Firms that have poor investment opportunities but nonetheless continue to investment substantial sums in We fixed capital are likely to be suffering from unresolved governance problems. thus create contrasting classes by sorting firm/year observations for which bottom 2/3 of their industry in a Q is in the given year, constmcting a separate sorting for firm/year observations for which investment spending (relative to the replacement value of the capital stock) is in the top 2/3 of their industry in a given year, and then forming the intersection of these two discount rate that Table For these Jensen firms, executives use a 340 basis points lower than that used by shareholders (row 4 of 2). The problems. that is sets. results presented in First, the Table 2 have two implications for governance economically and statistically significant gap between r and emerges from a variety of sortings strongly indicate the importance of govemance problems highlighted by the Free average of the eight estimated gaps between investment decisions for Jensen finns is r Cash Flow model. Based on the and r , the discount rate guiding 325 basis points lower than that used by r 16 shareholders." All of the estimates in Table 2 are statistically significant at conventional levels. Second, the Perfect Markets model holds that markets are largely able to overcome corporate governance problems and hence be any systematic gaps between r and considered here. This key implication 6. that there will not r based on the different classifications is decisively rejected in Table 2. Curing Free Cash Flow Problems? The results in Section 5 indicate that govemance problems associated with the Free Cash Flow model loom large for several classes of firms. Quantifying govemance problems in terms of discount rates presents an opportunity to evaluate the effectiveness of various corporate control costs of fi'ee cash flow. In his initial article emphasized that leverage is a increased debt loads borne mechanisms on in reducing the agency free cash flow, Jensen (1986) key financial control mechanism and thus by major American companies were that the beneficial as a corporate control device disciplining executives. While not discounting the benefits of leverage, Jensen's (1993) Presidential Address to the American Finance Association highlighted the role of strong institutions in curbing managerial abuses. These institutional control mechanisms can be assessed with our data because of the presence of interrelated groups and concentrated ownership for many Canadian firms. This section focuses exclusively on those firms suffering fi^om Free Cash Flow problems, and quantifies the extent to which these problems are attenuated by various The conclusion low market rates seems to stand in by surveying corporate executives (e.g., Blume, Friend, and Westerfield (1980, especially Tables 17 and 1 8) and Poterba and Summers (1995)). The latter authors received completed questionnaires from 228 U.S. fums and, among other that effective discount rates are relative to contrast to previous studies that have quantified discount rates findings, report that hurdle rates are "distinctly higher than equity holders' average rates of return and much higher than the return on debt during the past half-century" (p. 43). Apart from any differences in methodology (survey responses vs. actual spending patterns), these comparable to our results specifically on subsets of firms most because they apply to likely to all numbers are not firms, while our estimating strategy focuses be affected by a particular govemance problem. 17 control mechanisms. We construct the sample used in this section from the full sample of 193 fmns by identifying those firm/year observations with high free cash flow and poor investment opportunities. These are the same fmn/year observations discussed in Section 5 for which the estimates presented in row executives use a lower discount rate than shareholders. In this section, refer to these firm/year observations as "Jensen fimis." determines whether this discount rate gap (r - r ) for show 2 of Table 2 Our estimation we that simply strategy one class of Jensen firms differs systematically from that for a contrasting class of Jensen firms influenced by a particular control mechanism. We refer to this "difference-in-the-gaps" as ^. Contrasting classes of firms are formed by three control characteristics — debt, group membership, and concentrated ownership. Many authors have emphasized that increased debt loads could serve as a useful control device. " In principle, debt finance could curb the temptation to squander resources in several ways: by withdrawing cash from the hands of misbehaving executives, by increasing the danger of the firm in a period of financial distress, or markets where effective monitoring is that executives might lose confrol by forcing finns into external capital undertaken by knowledgeable outsiders. examine the effectiveness of debt as a control mechanism, we To classify firm/year observations as high-debt firms if their short-term and long-tenn debt nonnalized by the replacement value of fixed capital industry in a given year. If debt is (DEBT) is greater than the median an effective confrol mechanism, then, Jensen firms, higher debt should be associated with higher discount case, the gap between executives' and shareholders' discount for their among rates. rates should In this be reduced. Since ^ equals c F p ^ [(r High Debt Jensen " T High Debt Jensen) " (r Low Debt Jensen " T Low Debt Jensen)] ' Among other studies (1990), and and the gapS analyzing debt and control, see Grossman and Hart (1982), Harris and Raviv Aghion and Bolton (1992). 18 (r^ - r^) are negative, However, as shown -^ will be positive //debt row in 1 of Table is an effective control mechanism. very close to zero, and there 3, the ^'s are is no evidence suggesting that debt reduces governance problems. The second control mechanism we examine characteristic is based on an of the Canadian economy. Unlike the United institutional States, Canada has several major Canadian industrial groups involving ownership or other connections among many distinct enterprises with their own publicly traded shares. (Several countries have similar interrelated firms; see La Porta, Lopez-de-Silanes, and Shleifer (1999) for further discussion.) related to a their A firm is classified as having "group membership" if Canadian conglomerate own publicly-traded that brings together many it is distinct enterprises with 1 'K shares. Each firm typically through equity ownership. Many is effectively controlled by the group, of the groups are associated with a particular family or individual (e.g.. Black, Bronfrnan, Reichman), but not all (e.g., Bell Canada, Canadian Pacific). Group membership has uncertain effects as a governance mechanism. For firms in an interrelated group, long-term relations and fi^equent dealings the lead firm and affiliates and effective monitoring. conclusion. The among affiliates However, recent work theoretical analysis ~ arguably create casts doubts on this ~ between incentives for sanguine of Bebchuk, Kraakman, and Triantis (2000) highlights weaknesses with a cross-ownership structure as a control mechanism. Relative to executives in a dispersed ownership firm, the controllers of a firm with a cross-ownership structure are entrenched; relative to an owner with a very large equity stake, cash flow rights (and hence the incentive to create value) are low. Consequently, governance problems for group firms may be greater than for other firms. Data on group membership, as well as concentrated ownership discussed below, are obtained from Statistics Canada's Intercorporate Ownership 1984. 19 Among is firnis some evidence row 2 of Table 3, with high free cash flow and poor investment opportunities, there that group membership the gap is associated with better performance. between executives' and shareholders' discount rates In is about 200 basis points narrower for Jensen finns that are group members than for Jensen firms that are independent. However, as measured by statistically significant at conventional levels, a result that somewhat small sub-sample of Jensen firms belonging -^^ may to a the difference be partly due is not to the group with which the estimates are computed. Concentrated shareownership Canada, unlike the United main stock exchange 89) report that the third control mechanism we examine. in excess 58% of Canadian of 2 billion US In of the finns fraded on the States, a substantial proportion are majority-owned or controlled sample of firms with sales p. is by a single shareholder. For a dollars, Rao and Lee-Sing (1995, companies are concenfrated (per the above definition), while the proportion drops to 9% in the United States. Similar proportions are reported by La Porta, Lopez-de-Silanes, and Shleifer (1999, Table II.B). Nearly half of the firms in our full defined as either one shareholder having sample have concenfrated ownership 50% or more of the shares or as the finn being effectively controlled by another firm. (For example, Brascan was effectively confrolled by the Bronfhian family, although they only The problems role is of concenfrated ownership also ambiguous. By in owned 43% of the resolving corporate shares.) govemance internalizing the costs of exercising confrol, concenfrated ownership eliminates free-rider problems, and thus provides confroUing owners with the incentive For Jensen firms, to expend resources this additional to monitor monitoring would tend to reduce the gap between executives' and shareholders' discount rates, and thus ^. However, and discipline management. for substantial levels of concenfration, for the owner's private benefit, as suggested would be fmn reflected in a positive resources may be by the premium paid on large directed 20 blockholdings (Barclay and Holdemess, 1989). The adverse effects of concentration may be particularly acute in inherited. Canada because many large ownership stakes are Morck, Strangeland, and Yeung (2000) document the detrimental this particular effects of type of ownership (dubbed the "Canadian Disease") on firm performance. In this case, The empirical shareownership is we would expect ^ to be negative. results presented in row 3 of Table 3 indicate that concentrated associated with fewer governance problems. executives' and shareholders' discount rates shrinks The gap between by more than 300 basis points for Jensen firms with concentrated ownership (relative to Jensen firms with dispersed ownership), effectively eliminating the gap. These estimated ^'s are statisfically significant at the 7. 5% level. Capital Structure And Discount Rates In a recent survey of the literature, Myers (2001) lists the Free Cash Flow and Pecking Order models as two of the leading theories of capital structure. Does the evidence presented so far, which favors the Free Cash Flow model, preclude the empirical importance of the financing problems highlighted by the Pecking Order model? Not necessarily, since some firms may be constrained by a Pecking Order while others face Free Cash Flow problems. may vary over time. For example, firm may have projects, Even for a given firm, these problems with unexpected variation in product demand, a insufficient inside funds to finance its positive net present value and face relafively higher costs when securing outside finance. At another point in time, the same firm may find itself with modest investment opportunities, surplus cash flow, and a substantial temptation to squander resources. Such time- varying changes in circumstances are at the core of Stulz's model of capital structure Myers (2001 , p. 81 ), specifically argues that firms that are prone to overinvest." "The free cash flow theory is designed for mature 21 in which the relative costs mix of debt and equity estimated from a of governance and finance problems determine the optimal (Stulz, 1990). common framework financial policy and, more Measuring these costs in tenns of discount rates generates insights into this theory of optimal generally, the factors affecting capital stnicture.'^ Higher costs of outside finance arise when suppliers of extemal finance cannot assess the quality of firms and their investment projects (Myers and Majluf, 1984). Given an inability to discriminate between high and low quality firnis, '^ shareholders add a "lemons premium" to the cost of outside finance. Adverse selecfion occurs because high quality firnis tend to avoid extemal capital markets, reducing the quality of the pool of borrowers and creating a sustainable positive difference between the costs of inside and outside funds. This difference creates a financial hierarchy or Pecking Order (in which firms exhaust inside fiinds before using relatively expensive outside fiinds), and depends on the limitations firm face in accessing extemal capital markets. This section focuses on those firms most likely to both need outside finance and face a relafively higher cost for outside fiinds (thereby using a higher effective discount rate) as predicted by the Pecking Order model. financing problems is Our strategy for assessing similar to that for governance problems: we identify a subset of firms likely to face finance constraints and, based on their pattem of investment spending, estimate the difference in discount rates between outside and inside finance. We use two characteristics to define contrasting classes of firm/year observations. Firms with good investment opportunities are more likely to exhaust intemal sources of finance and rely more heavily on extemal sources where financing problems The may close relation between governance and finance problems has been noted in the survey of and Vishny (1997, p. 773): "Corporate governance deals with the agency problem: the separation of management and finance." Shleifer See Hubbard (1992), and Ng ( 1 998) for a recent survey and the work of Hubbard and Kashyap and Schaller (1996) for related analyses. ( 1 992), Whited 22 arise. We identify these firms by sorting the firm/year observations for which Q is in the top 2/3 of their industry in a given year. The second characterisfic is based on the higher financing costs faced by firms with hmited access to external capital markets. Limited access by an is measured in three ways: by the absence of a bond rating (measured indicator variable taking the value of one if the firm does not have a rating, zero elsewise; data are obtained Dominion Bond Ratings the highest values of bond from the Canadian Bond Ratings Service and by the two-thirds of firm/year observations with Service), DEBT (short-term and long-term debt normalized by the replacement value of fixed capital) in their industry in a given year, or by the two-thirds of the smallest firms (measured by SIZE defined in terms of the replacement value of the fixed capital stock). Firms that are more likely to require high-cost outside finance are defined by the intersection of firm/year observations with good investment opportunifies and limited access to external credit markets, the latter measured by one of the three preceding character-istics. In each case, we contrast these firms with the remaining firms in the sample. Table 4 contains the results for sortings by finance problems. In all three cases, the results clearly indicate the presence of economically important financing problems. All estimates of differences (labeled p) between the cost of outside and inside funds are statistically significant at the discount rates for outside and inside funds is 1% level. The difference between the estimated at 900 basis points when we focus on the lack of a bond rafing as a measure of the difficulty of accessing outside funds, 760 basis points when we when we focus on high focus on smaller firms. If high quality Canadian fiims prefer levels of debt, and 910 basis points As with Free Cash Flow governance problems, we to obtain funds from the larger US market, these firms might bond ratings only fi^om US rating agencies. In this case, our procedure, which is based on ratings by Canadian rating agencies, will misclassify these high quality firms with ready access to external capital markets. There are no firms in our sample that are not rated by a Canadian rating agency but are rated by S&P or Moody's. Thus, there does not appear to be a flight of high quality obtain borrowers to the larger US market and its rating agencies. 23 find strong evidence that financing problems are economically These significant. results suggests that, consistent with the capital structure choices are and statistically model of Stulz (1990), detennined by a mixture of governance and finance problems whose applicability varies by firm and, for a given firm, across time. 8. Summary And Conclusions new This paper sheds light on long-standing corporate governance use information from a transformed NPV rule issues. We and variation from finn-level panel data to uncover effective discount rates. In this revealed preference approach, effective discount rates guiding investment decisions are inferred from the pattern of investment spending. The gap between the discount rates used by executives and shareholders is our measure of governance problems. The empirical work is based on panel data for 193 Canadian firms. Distinctive institutional features, such as interrelated groups of Canadian firms and concenfrated share ownership, allow us to quantify the sensitivity of discount rate gaps and governance problems to these institutional control Our key mechanisms. finding is Cash Flow model prove results, executives that the to governance problems highlighted be economically and in Jensen's statistically significant. Among other of firms with high levels of free cash flow and poor investment opportunities use effective discount rates that are 350-400 basis points below share- holders' discount rates. Moreover, the key implication of the Perfect Markets - Free that there should not be any systematic bias in discount rates ~ is model - decisively rejected. With model, we this documentation of the problems associated with the Free Cash Flow then investigate possible cures for Jensen firms ~ those with high free cash flow and poor investment opportunities. In our data, neither debt nor membership an interrelated group of finns has any in statistically significant effect in attenuating free 24 cash flow problems. However, the gap between executives' and shareholders' discount rates is effectively eliminated for Jensen firms with concentrated share ownership. These results suggest that institutional arrangements can be effective in curbing corporate control problems, and are consistent with Shleifer and Vishny's (1997, p. 769) conclusions concerning the prominent role played by large shareholders in promoting good corporate governance. Lastly, we use our revealed preference capital structure models. For some firms some at important. At other times, perhaps for the approach to consider two prominent same times, Free Cash Flow problems are Pecking Order problems are firm. important. Thus, explanations of capital structure choices will have to rely on several theories depending on circumstances that vary across firms and, for a given firm, across time. Apart fi-om evaluating hypotheses, the parameter estimates permit us to quantify the impact of corporate governance problems on capital accumulation. Jorgensonian user cost of capital is The the relevant price variable determining capital accumulation, and depends on the discount rate guiding investment decisions. When executives use a discount rate that is below that of shareholders, the Jorgensonian user cost will be distorted. Combining the Jorgensonian user cost with an assumption about the price extent to I elasticity of the demand for which Free Cash Flow problems lead firms Q The Jorgensonian user cost of capital is defined as follows: to hold too ) or executives' (r ) can estimate the much UC = (r+5)*p F S rate (either the shareholders' (r we capital, fixed capital. T , where r is the discount • discount rate), 5 is I the depreciation rate, and the price of investment goods relative to the price of output, adjusted for corporate income p is taxes, investment tax credits, and the present value of tax depreciation allowances (Hall and Jorgenson, 1971). Intuitively, this user cost is the rate for renting a capital good for one period. This rental rate good value of economic depreciation by 5 includes a "nonrefundable security deposit" reflecting that only a fraction of the rented capital will be returned because in the user cost capital formula of depreciation is (5). Representing the equivalent to calculating the present value of a stream of deductions for a good depreciating according to a declining-balance formula at rate 5. 25 With executives using discount rates 325 basis points lower than shareholders arithmetic average of the estimates in Table 2), the user cost of capital falls If the price elasticities of capital range from -0.25 to -0.75, firms suffering unresolved governance problems have approximately Our conclusions for some firms Whether these at some are that Free times, (the by 29.5%. from 7% to 22% too much capital. '^ Cash Flow problems are quantitatively important and can be ameliorated by concentrated ownership. results extend to other countries, other legal frameworks, and other institutional settings are topics for future research. ] Q The percentage change (AK/K) equals the percentage change in the Jorgensonian user cost of capital (aUCAJC) multiplied by the price elasticity (-9). aUC/UC = (r r^)*p' / (r^+S)*p' = -0.0325 / (0.032 + 0.080) = -0.290, where the value for is obtained from in the capital stock rW" Table and 6 equal to 0.080 is an average of our estimates of firm-specific depreciation rates computed with firms' reported depreciation and the procedure in Salinger and Summers (1983). AK/K = aUC/UC * (-9) = -0.290 * (-0.50) = 0.145 based on a value of 9 of 0.50. 1 26 References Aghion, Philippe, and Bolton, Patrick, "An Incomplete Contracting Approach to Financial Contracting," The Review of Economic Studies 59 (July 1992), 473-494. Alchian, Armen, "Uncertainty, Evolution, and Economic Theory," Journal of Political Economy 58 (1950), 21 1-221. Barclay, M.J., and Holdemess, C.G., "Private Benefits from Control of Public Corporations," Journal ofFinancial Economics 25 (1989), 371-395. Bebchuk, Lucian, Kraakman, Reinier, and And Dual Triantis, George, "Stock Pyramids, The Mechanisms And Agency Costs Of Separating Control From Cash Flow Rights," in Randall K. Morck (ed.). Cross-Ownership, Class Equity: Concentrated Corporate Ownership (Chicago: University of Chicago Press (for the NBER), 2000), 295-315. Blume, Marshall E., Friend, Irwin, and Westerfield, Randolph, Impediments Capital Formation (Philadelphia: Rodney L. White Center for Financial Research, to 1980). J., "Explaining the Shape of Corporate Governance Law: The Managerial and Decision Economics 18 (November/ Role of Competition," Carney, William December 1997), 61 1-626. Chirinko, Robert S., "Business Fixed Investment Spending: Modeling and Policy Implications," Journal ofEconomic (December 1993), 1875-191 1. Strategies, Empirical Results, Literature ?,\ Daniels, Ron, and Morck, Randall (eds.). Corporate Decision-Making In Canada Industry Canada/Financial Research Foundation of Canada Research Volume V (Vancouver: University of British Columbia Press/University of Calgary Press, 1995). Demsetz, Harold, and Lehn, Kenneth, "The Structure of Corporate Ownership: Causes and Consequences," Journal ofPolitical Economy 93 (December 1985), 1155-1177. Dow, James, and Gorton, Gary, "Stock Market Efficiency: Is There a Connection?," Journal Efficiency and ofFinance 52 Economic (July 1997), 27 1087-1130. 28 Easterbrook, Frank, and Fishel, Daniel, The Law (Cambridge: On Economic Structure of Corporate Harvard University Press, 1991). Fama, Eugene and French, Kenneth R., "Common Risk Factors In The Returns Stocks and Bonds," Journal of Financial Economics 33 (1993), 3-56. Grossman, Sanford, and Hart, Oliver, "Corporate Financial Structure and Managerial Incentives," in J. McCall (ed.). The Economics of Infonnation and Uncertainty (Chicago: University of Chicago Press, 1982), 123-155. Bronwyn and Hall, Robert E., "The Value and Perfonnance of U.S. Corporations," Brookings Papers on Economic Activity (1993: 1), 1-50. Hall, H., and Jorgenson, Dale W., "Application of the Theory of Optimum Capital Accumulation," in Gary Fromm (ed.). Tax Incentives and Capital Hall, Robert E., Spending (Washington: Brookings Institution, 1971), 9-60. Hansen, Lars Peter, "Large Sample Properties of Generalized Method of Moments Estimators," Econometrica 50 (July 1982), 1029-1054. and Raviv, Artur, "Capital Structure and the Information Role of Debt," Journal ofFinance 45 (1990), 321-350. Harris, Milton, Hart, Oliver, "Corporate Governance: Some Theory and Implications," The Economic Journal 105 (May 1995), 678-689. Hatch, James E., and White, Robert W., Canadian Stocks, Bonds, Inflation: 1950-1987 (Charlottesville, Virginia: Bills, Research Foundation of the and Institute of Chartered Financial Analysts, 1988). Hayashi, Fumio. "Tobin's Marginal q and Average q: Interpretation." Econometrica 50 (January 1982), 213-224. A Neoclassical Hubbard, R. Glenn, "Capital-Market Imperfections and Investment," Journal of Economic Literature 2)6 (March. 1998), 193-225. Hubbard, R. Glenn, and Kashyap, Anil, "Internal Net Worth and the Investment Process: An Application to U.S. Agriculture, 1910-1987," Journal of Political Economy 100 (June 1992), 506-534. . 29 C, "Agency Costs of Free Cash Flow, Corporate Takeovers," American Economic Review 76 (May 1986), 323-329. Jensen, Michael Jensen, Michael C, "The Modem Industrial Revolution, Exit, Finance, and and the Failure of Internal Control Systems," Journal of Finance (July 1993), 831-880. La Porta, Rafael, Lopez-de-Silanes, Florencio, and Shleifer, Andrei, "Corporate Ownership Around the World," Journal of Finance 54 (April 999), 471-51 8. 1 American Corporate Governance Fatally Flawed?," in Donald H. Chew (ed.). Studies In International Corporate Finance and Governance Systems: A Comparison Of The U.S., Japan. & Europe (New York and Oxford: Oxford University Press, 1997), 38-45. Miller, Merton H., "Is Morck, Randall K., Stangeland, David A., and Yeung, Bernard, "Inherited Wealth, Corporate Control and Economic Growth: The Canadian Disease," ," in Randall K. Morck (ed.). Concentrated Corporate Ownership (Chicago: University of Chicago Press (for the NBER), 2000), 319-369. Myers, Stewart C, "Capital Structure," Jowr«a/ Of Economic Perspectives 15 (Spring 2001), 81-102. Myers, Stewart C, and Majluf, Nicholas Investment Decisions S., "Corporate Financing and When Firms Have Information That Investors Do Not Have," Journal of Financial Economics 1 3 (June 1 984), 1 87-22 1 Ng, Serena, and Schaller, Huntley, "The Risky Spread and Monetary Policy Transmission: Evidence on the Role of Asymmetric Information," Review OfEconomics And Statistics 78 (1996), 375-383. James M., and Summers, Lawrence H., "A Survey of U.S. Companies, Time Horizons, and Hurdle Rates," Sloan Management Review 37 (1995), 43-53. Poterba, Someshwar, and Lee-Sing, Clifton R., "Governance Structure, Corporate Decision-Making and Firm Performance in North America," in Ron Daniels and Randall Morck (eds.). Corporate Decision-Making In Canada Industry Canada/Financial Research Foundation of Canada Research Volume V (Vancouver: University of British Columbia PressAJniversity of Calgary Press, 1995), 43-104. Rao, P. 30 Romano, Roberta, The Genius ofAmerican Corporate Law (Washington: American Enterprise Institute Press, 1993). Salinger, Investment: Michael A., and Summers, Lawrence H., "Tax Reform and Corporate A Microeconomic Simulation Study," Behavioral Simulation Methods Chicago Press (for the NBER), in in Martin S. Feldstein (ed.). Tax Policy Analysis (Chicago: University of 1983), 247-273. and Vishny, Robert W., "A Survey of Corporate Governance," Journal of Finance 52 (June 1997), 11)1-1^?>. Shleifer, Andrei, Jeremy C, "Rational Capital Budgeting In Journal ofBusmess 69 (October 1996), 429-455. Stein, Stulz, An Irrational World," Rene, "Managerial Discretion and Optimal Financing Policies," Journal ofFinancial Economics 26 (July 1990), 3-27. Whited, Toni M., "Debt, Liquidity Constraints, and Corporate Investment: Evidence from Panel Data," Journal ofFinance 47 (September 1992), 1425-1460. 31 Appendix A Departures To From Value Maximization further motivate our revealed preference approach, this Appendix examines the imphcations for discount rates of departures from value maximization maximizing executives. between the discount by utility A parsimonious model links governance problems to the gap rates used by executives and shareholders. The version of the Free Cash Flow model presented here emphasizes the utility that executives receive from empire-building. Consider the following model of a Jensen firni, defined as a firm where executives obtain utility from both shareholder value and the size of the firm. Profit (tt) 71 equals revenues (f[K,L]) less labor costs (w*L) less capital rental costs ((r+5)*K), = f[KX] - w*L where fIK,L] is wage the is rate. - (A-1) (r+5)*K, a production function depending The user cost of funds and 5 is on capital (K) cost of capital equals (r+5), where the rate of depreciation (see footnote 1 and labor (L), and w r is the real risk-adjusted 8 for further discussion of the user cost of capital). (The roles of taxes and relative prices have been omitted.) For the standard neoclassical firm, the user cost of capital rate. Consistent with the Free Cash U = is the appropriate discount Flow model, executives obtain and some other characteristic of the firm, such as Jensen firm is size. The utility from profit objective flinction for a as follows, U[7r,K]. We specify U[.] as a Cobb-Douglas fiinction, which provides a succinct way of representing the differing weights placed on profit (0<P<1) and size (0<y<1), (A-2) 32 U = The /*K\ first-order condition for utility Jensen firm = (A-3) . is maximization with respect to the choice of K by a as follows, p*/''*K^*(fK[K,L]-(r+5)) + j*n^*K'''\ (A-4) we Rearranging (A-4), capital obtain the following equafion for the marginal product of (MPK / fK[K,L]) that implicitly MPK^ = MPK^ + determines the optimal capital stock, (A- (j), 5a) (|) = _(y/p)*(7r/K)<0. (A- 5b) where MPK and MPK are the MPK's determined by optimizing behavior by Jensen and standard neoclassical firms, respectively. The key relations alternative utility functions. replaced by a CES When the Cobb-Douglas (with substitution elasficity r|) in (A-5) are robust to specificafion in (A-3) or a linear function with is % and K as arguments, the ^ in equation (A-5b) becomes -(y/p)*(7r/K)'^ and -(y/p), respectively. There are several interesting features to MPK from adjustments for depreciation and risk, shareholders, and hence ^ in (A-5a) equal to is this (r model highlighted by (A-5). Apart is - r the discount rate required ). If the firm is operating in a perfectly competitive environment with constant retums to scale (7r=0), then and we obtain the by (j) = 0, standard neoclassical relation equating the marginal product of capital to its user cost. This result reflects that Jensen-type exist in equilibrium only if some rents accrue. govemance problems can The discount rate gap will also be zero 33 if executives receive utility utility only from profit (y=0) or if executives obtain perquisite only from variables not entering the value maximization problem, such as an honorary seat on a local arts council. However, such a seat is more likely to be offered the larger the size of the firm. For the Jensen fimi, the standard neoclassical relation fails to hold because executives obtain perquisite utility (y>0) in a non- competitive environment (7i>0). F S estimates of (r -f). It is precisely these effects that are captured by our 34 Appendix B The Real Risk-Adjusted Market Discount And Data Variable Construction The discount factor (Ri,t, market discount rate defined = r,,t Rate: Sources defined in (9)) depends on the real, risk-adjusted in the text ( 0) as follows, 1 (Bl) ((l+i,,t)/(l^'t))-1.0. This rate discounts annual cash flows fi"om the middle-of-period period t+1 . The equity risk three-factor (FF3) models. premium estimated by the is The components of rt t to the middle-of- CAPM or the Fama-French are defined and constructed as follows, Nominal, short-term, ii,t risk adjusted company the middle-of-period (MOP) t to the . N -DEBT X -EQUITY y-i /I ^(1-T,,t)l + (l-?w)l i.t. cost of capital fi^om MOP t+1, -, , t • 1 DEBT Nominal, one year, Commercial Paper t MOP t+1 . This rate is rate fi-om the MOP t constructed fi-om monthly data for the Canadian 90-day Commercial Paper Rate, average of daily (CANSIM B 140 17). series to the into beginning-of-period rates These monthly data are converted (BOP) monthly data for January, April, and October by averaging the monthly data in the preceding and current months. For example, the January BOP rate is the arithmetic average of the monthly December and January rates. July, TO. The and nr\Tt BOP rates are representedJ as i i 4. t.ocr- The one year -DEBT -DEBT -DEBT tjuL, uan, i iapr? i 1 MOP rate is constructed fi:om these BOP rates as follows, r/1 • I DEBT [(1+1 \ /I • DEBT \ /i I t,JUL)(l+l • , t,OCT)(l+l DEBT a /i t+ 1. J • , an) (1+1 DEBT nt.25 t+l,APR)] - 1. •EQUTTY 1 i.t Nominal, short-term, MOP •F + t to the a,. risk adjusted cost MOP t+1. of equity capital fi-om the 35 i "= t Nominal, one year, the MOP t+1 . risk free This rate is Treasury Bill rate from the MOP t to constructed from monthly data for the Canadian 90-day Treasury average yields Bill Rate, at weekly (CANSIM series B 14007). These monthly data are converted into BOP monthly data fr)r January, April, July, and auctions October by averaging the monthly data in the preceding and current months. For example, the January BOP rate is the arithmetic average of the monthly December and January rates. The BOP rates are represented as i.jan, i.apr, t,juL, and t.ocTThe one year MOP rate is constructed from these BOP rates as i i i i follows, [(1+UjUl) = 71^ (l+iWl) (l+iVl,JAN) (1+i^+l.APR)]^^ Oneyearinflationratefrom theMOP constructed from monthly All Items (CANSIM into series theMOPt+1 Price Index data P700000). These monthly data are converted BOP monthly data for July by inflation rate is constructed (CPIt.,,JUL/CPI,JUL) = to Consumer averaging the monthly data in June and July, and are represented as a t - 1. - CPIt.juL- from these The one year BOP rates MOP as follows: 1.0. Equity risk premium. The methods used to estimate a, are discussed below. = Ti,t Marginal rate of corporate income taxation (federal and provincial) incorporating variation across time and industries. = Xi Finn specific leverage ratio calculated as the time-average of bonds (bonds + equity), where bonds are at book value and equity at market value. Two methods Under the CAPM, are used to estimate the equity risk premium, a. / 36 O, = n EQUITY / Fx (B2) - |i ), Pi (|I where CAPM from Hatch and White (1988, Table 5-36). These industry P's are assigned to firms on the basis of the Toronto Pi (3 Stock Exchange's industry classification. EQUITY The source Total return on equities from 1950-1986: 0.1 148. is Hatch and White (1988, Table 5-15). F Total return on risk free Treasury bills from 1950-1986: 0.0584. The source Hatch and White (1988, Table is 5-5). We also measure the equity risk premium using the Fama and French (1993) three-factor a, = where model (FF3), n EMR [i EMR |i p, is + , r, p, SMB SMB |i + , n HML a mathematical expectation, HML (B3) \i Pi EMR is the excess market retum (value-weighted market retum minus risk-free rate), HML is the book-to-market risk factor, P factor loadings The on these three p , , SMB is the size risk factor, and and P risk factors. portfolios are constructed as follows: firms a given year if: 1) are the firm-specific book equity (common stock available) and market equity for the were included capital, plus deferred in the sample in income taxes end of the previous year were available if in the Financial Post dataset; and 2) returns data for the current year were available from the Toronto Stock Exchange (TSE)-Westem dataset. All firms in the sample year were ranked on size (using market equity) and split into in a given small and big (S and B) depending on whether they were above or below the median. All firms in the sample 37 in a given year were then ranked by the ratio of book equity to market equity with breakpoints for the bottom 30% (Low), middle Six portfoHos (S/L, S/M, S/H, B/L, 40% (Medium), and top B/M, B/H) were constructed from 30% (High). the intersection of the two size and three book-to-market categories, and monthly value-weighted returns on the six portfolios were calculated for each calendar year. SMB is defined as the difference, each month, between the simple average of the returns on the three small firm portfolios (S/L, S/M, and S/H) and the corresponding simple average of the returns on the three big firm portfolios. HML is defined as the difference, each month, between the simple average of the returns on the two high-book-to-market portfolios and the corresponding simple average of the retums on the two low-book-to-market portfolios. The i factor loadings on the three were estimated from a regression of excess retums risk factors at a monthly frequency, [i ,\i annualized geometric means of the three risk factors over 1986 inclusive. , all and \i for finn are the months from 1973 to 38 Table 1 Summary The Statistics entries are the median, [mean], and (standard deviation) of the indicated variable calculated from 1975 to 1985 for 193 Canadian firms. COST is variable costs (sales minus operafing income) divided by K$, the current dollar replacement cost of the fixed capital stock. DEBT is short and long term debt divided by K$. FCF] is free cash flow defined as cash flow (revenues less cash expenses including tax and interest payments) less dividends (all for the previous year) divided by K$. FCF2 is an altemative definition of free cash flow defined as cash flow less dividends (averaged two years) all divided by K$. INVT is gross capital expenditures on property, plant and equipment divided by K$. p is the tax-adjusted price of investment goods relative to the price of output. Q is the Brainard-Tobin Q, the financial value of the firm less the tax depreciation bond less K$ (adjusted for taxes) all divided by the fixed capital stock valued in terms of the price of output (see footnote 10 for details), r andr are the real risk-adjusted market discount rates defined in Section 4.2 and Appendix B with the equity risk premia estimated by the CAPM and Fama-French three-factor (FF3) models, respectively. REV is revenues divided by K$. SIZE equals K$ stated in millions of Canadian dollars. N is the number of firm/year observations. Column 1 is for all firms in the sample. Columns 2 and 3 are sorted by whether a firm's primary operations are in the oil industry, TSE Industry 300; columns 4 and 5 are sorted by whether fimi/year observations for Free Cash Flow (FCFi) are in the top 2/3 of their industry in a given year and firm/year observations for investment opportuniries (Q) are in the bottom 2/3 of their industry in a given year. For fiirther details about data definitions and the sorting variables, see Sections 4 and 5, respecfively. for the previous 39 Table 1 Summary Statistics (Continued) All Firms High Free Cash Flow Oil Industry & Poor Investment Opportunities COST DEBT FCF, FCF2 INVT P^ Yes No Yes No (1) (2) (3) (4) (5) 2.406 1.223 2.609 2.174 2.538 [4.195] [2.993] [4.382] [3.575] [4.656] (6.096) (5.546) (6.158) (4.484) (7.026) 1.001 0.671 1.046 0.809 1.258 [1.559] [1.228] [1.611] [1.074] [1.920] (1.978) (1.888) (1.987) (0.971) (2.411) 0.071 0.071 0.071 0.083 0.055 [0.069] [0.076] [0.068] [0.090] [0.054] (0.063) (0.054) (0.064) (0.039) (0.073) 0.067 0.069 0.067 0.078 0.055 [0.067] [0.072] [0.066] [0.083] [0.056] (0.053) (0.044) (0.055) (0.038) (0.060) 0.170 0.201 0.166 0.151 0.187 [0.248] [0.260] [0.246] [0.192] [0.289] (0.350) (0.224) (0.366) (0.205) (0.423) 0.952 0.955 0.951 0.965 0.940 [0.979] [0.986] [0.979] [0.988] [0.973] (0.168) (0.171) (0.168) (0.166) (0.170) 40 Table 1 Summary Statistics (Continued) All Firms High Oil Industry Freie Cash Flow & Poor (1) Q CAPM r r^""^ REV SIZE N Inv(jstment Opportunities Yes No Yes No (2) (3) (4) (5) -0.905 0.203 -0.409 -0.128 -0.455 [0.688] [0.287] [0.750] [-0.897] [1.865] (5.911) (1.897) (6.309) (1.651) (7.460) 0.040 0.050 0.036 0.043 0.038 [0.040] [0.049] [0.039] [0.037] [0.043] (0.036) (0.034) (0.036) (0.033) (0.038) 0.032 0.030 0.041 0.027 0.035 [0.035] [0.043] [0.033] [0.031] [0.037] (0.039) (0.039) (0.039) (0.037) (0.040) 2.834 1.474 3.067 2.510 3.059 [4.733] [3.373] [4.945] [3.964] [5.304] (6.577) (5.987) (6.640) (4.719) (7.625) 62.783 80.438 60.575 65.147 61.129 [323.956] [466.056] [301.833] [370.240] [289.567] (689.773) (989.789) (627.847) (783.628) (608.836) 2123 286 1837 905 1218 41 Table 2 Assessing Governance Problems GMM estimates of equation (8); discount factor defined in equation parameters in columns and 1 3 are estimated (9). The based on the Fama-French three-factor (FF3) estimate of the equity risk premium; cokmins 2 and 4 are based on the CAPM estimate of the equity risk premium. Contrasting classes are defined in the rows as row firms with their primary operations in the oil industry (TSE industry 300); row 2, firm/year observafions for which Free Cash Flow (FCFi) is in the top 2/3 of their industry in a given year intersected with the finn/year observations for which follows: 1 , investment opportunities (Q) are in the bottom 2/3 of their industry in a given year; row 3, the same procedure as used in row 2 with FCFi replaced by Averaged Free Cash Flow (FCF2); row 4, the same procedure as used in row 2 with FCFi replaced by the investment/capital ratio (INVT). Variables are defined in Table and Appendix B for more extensive definitions, (r - r ) = (j) is 1 ; the gap see Section 4 between the discount rates used by executives and shareholders for firms indicated by the label. \\i is the parameter capturing deviations compefifion. A value of v|; from constant returns less than unity is consistent row to scale or perfect with decreasing returns to scale regardless of the competifive nature of output markets or increasing returns and a sufficient degree of imperfect competition to force the marginal return to capital below its economic average return. In either case, rents. The instruments it implies that the firnis are earning positive are listed in footnote 8. Standard errors are in parentheses. ***, **, and * indicate statistical significant at the 1%, levels, respectively. 5%, and 10% 42 Table 2 Assessing Governance Problems (Continued) Estimated Parameters Contrasting Classes Gap Between And Executives' Measure Of Economic Rents Shareholders' Discount Rates (r'-r^) = ¥ (l) FF3 (1) 1 . 2. CAPM (2) -0.026* -0.028* (0.014) (0.014) High Free Cash Flow^ -0.035** -0.040*** & (0.015) (0.015) High Averaged Free Cash Flow -0.034** -0.040*** & (0.015) (0.015) Oil Firms _EE3 (3) CAPM (4) 0.955*** 0.956*** (0.004) (0.004) 0.953*** 0.954*** (0.004) (0.004) Poor Investment Opportunities 3. 0.953*** 0.954*** (0.004) (0.004) Poor Investment Opportunities 4. High Investment -0.034*** -0.023* & (0.013) Poor Investment Opportunities (0.013) 0.953*** 0.954*** (0.004) (0.004) 43 Table 3 Curing Free Cash Flow Problems? GMM estimates of equation (8); discount factor defined in equation (9). parameters in columns and 1 3 are estimated based on the Fama-French The tliree-factor (FF3) estimate of the equity risk premium; coluinns 2 and 4 are based on the CAPM estimate of the equity risk premium. All of the results in this table are based on the following "Jensen" sub-sample of finn/year observations constnjcted from the sample: the firm/year observations for which Free Cash Flow (FCFi) is full in the top 2/3 of their industry in a given year intersected with the firm/year observations for which investment opportunities (Q) are in the bottom 2/3 of their industry in a given year. Based on this Jensen sub-sample, contrasting classes are defined in the rows as follows: median row 1 , firm/year observations for which debt for their industry in a given year; that brings together row 3, many row 2, (DEBT) is greater than the firms related to a Canadian group distinct enterprises with their firms that either have one shareholder having own 50% or publicly-traded shares; more of the shares or that by another finn. Variables are defined in Table 1 see for more extensive definitions, y^ is the parameter measuring the difference between (r - r ) for one class of Jensen firms influenced by the control mechanism indicated by the row label and (r - r' ) for a contrasting class of Jensen firms Vv/ithout the control mechanism, y is the parameter capturing deviations from constant returns to scale or perfect competition. A value of vj/ less are effectively controlled Section 4 and Appendix than unity is ; B consistent with decreasing returns to scale regardless of the competitive nature of output markets or increasing returns and a sufficient degree of imperfect competition to force the marginal return to capital below its average return. In either implies that the firms are eaming positive economic rents. The instruments are listed in footnote 8. Standard errors are in parentheses. ***, **, and * indicate case, it stafistical significant at the 1%, 5%, and 10% levels, respectively. 44 Table 3 Curing Free Cash Flow Problems? (Continued) Estimated Parameters Contrasting Classes Reduction In The Gap Between Executives' And Measure Of Economic Rents Shareholders' Discount Rates -¥- _EE3_ (1) 1. 2. 3. High Debt Group Membership Concentrated Ownership CAPM (2) -0.006 -0.005 (0.022) (0.020) 0.021 0.019 (0.023) (0.024) 0.048** (0.020) 0.039** (0.020) JEE3 (3) CAPM (4) 0.963*** 0.963*** (0.006) (0.006) 0.963*** 0.962*** (0.006) (0.006) 0.963*** 0.963*** (0.006) (0.005) 45 Table 4 Capital Structure And Financing Problems GMM estimates of equation (8); discount factor defined in equation parameters in columns 1 and 3 are estimated (9). The based on the Fama-French three-factor (FF3) estimate of the equity risk premium; columns 2 and 4 are based on the CAPM estimate of the equity risk premium. Contrasting classes are defined in the rows as row which investment opportunities (Q) are in the top 2/3 of their industry in a given year intersected with the finns that do not have a bond rating; row 2, the same procedure as used in row 1 with the absence of a bond rating replaced by the firm/year observations for which debt (DEBT) is in the top 2/3 of their industry in a given year; row 3, the same procedure as used in row 1 with the absence of a bond rating replaced by the firms for which size (SIZE) is in the see Section 4 and bottom 2/3 of the sample. Variables are defined in Table Appendix B for more extensive definitions, (r - r ) = p is the difference between the discount rates on outside and inside fijnds for firms indicated by the row label. is follows: 1, the firm/year observations for 1 ; \\i the parameter capturing deviations fi^om constant returns to scale or perfect competition. A value of v|/ less than unity is consistent with decreasing returns to scale regardless of the competitive nature of output markets or increasing returns and a sufficient degree of imperfect competition to force the marginal return to capital below its average return. In either case, it implies that the firms are earning positive economic rents. The instruments are listed in footnote 8. Standard errors are in parentheses. ***, **, and * indicate statisfical significant at the 1%, 5%, and 10% levels, respectively. 46 Table 4 Capital Structure And Financing Problems (Continued) Estimated Parameters Contrasting Classes Difference Between Discount Rates Outside And On Measure Of Economic Rents Inside Funds ,0 (^ 1 - »• _EE3 (1) 1 . No Bond Rating 0.090*** . ) =p CAPM (2) 0.095*** ¥ _EE3 (3) CAPM (4) 0.954*** 0.954*** & Good 2. (0.017) & 3. 0.076*** (0.016) (0.004) 0.074*** (0.016) 0.956*** 0.957*** (0.004) (0.004) Investment Opportunities Small Size & Good (0.004) Investment Opportunities High Debt Good (0.016) 0.091*** (0.016) Investment Opportunities 0.089*** (0.016) 0.956*** 0.957*** (0.004) (0.004) AbstractSubmission@ssm.com, 03:02 PM 12/5/2002 -0500, Your SSRN Abstract Submis,..j Page 1 of 1 From: <AbstractSubmission@ssm.com> Subject: Your SSRN Abstract Submission Date: Thu, 05 Dec 2002 1 5:02:20 -0500 To: <iwoodbur@mit.edu> Dear Linda Woodbury: "A Reealed Preference Approach to Understanding Governance Problems: Evidence from Canada" has been submitted corporate Your abstract titled for publication in SSRN's abstracting journals. If you do not hear from a SSRN representative within 4 weeks asking for your approval of the abstract and related information to appear in our joumal(s), please email AbstractSubmission@SSRN.Com to inquire about its status. Please include the date abstract, the abstract, author, Thank you for your interest and paper in SSRN title in you submitted the your inquiry. publications. Sincerely, Michael C. Jensen Printed for Linda Woodbury <lwoodbur@mit.edu> 12/5/2002 7A$/^^ Date Due ' MIT LIBRARIES 3 9080 02528 5989 piOHJliSiT